Make Simulated Neurons

%load_ext autoreload
%autoreload 2
%matplotlib inline

from glob import glob
import matplotlib.pyplot as plt
import numpy as np
import os
import pandas as pd
import plotly.express as px
import sys
import tifffile as tf
import torch

sys.path.append(os.path.expanduser('..'))

from environments.sac_tracking_env import Environment
import json
from data_prep import generate, draw, load

Generate simulated SWC file data.

This is a list of nodes, each node being a list: [sample_idx, structure_id, x, y, z, radius, parent_id]

swc_list = generate.make_swc_list((101,101,101),
                        length=20,
                        step_size=3,
                        kappa=20.0,
                        uniform_len=False,
                        random_start=True,
                        rng=None,
                        num_branches=1) # make simulated neuron paths.

Set random background and foreground (neuron) colors

random_contrast=True
if random_contrast:
    neuron_color = np.random.rand(3)
    neuron_color /= np.linalg.norm(neuron_color)
    background = np.random.rand(3)
    background = background / np.linalg.norm(background) * 0.01

Draw neurons from SWC format data; one without and one with added noise and artifacts.

neuron_no_artifacts = draw.neuron_from_swc(swc_list,
                                width=3,
                                noise=0.0,
                                adjust=False,
                                neuron_color=None,
                                background_color=None,
                                random_brightness=False,
                                dropout=False,
                                binary=False)

neuron_with_artifacts = draw.neuron_from_swc(swc_list,
                                width=3,
                                noise=0.05,
                                adjust=False,
                                neuron_color=neuron_color,
                                background_color=background,
                                random_brightness=True,
                                dropout=True,
                                binary=False)

Draw the images.

a. Without artifacts

img = neuron_no_artifacts["image"].data.permute(1,2,3,0)

print(img.shape)
fig, ax = plt.subplots(1,3)
ax[0].imshow(img.amax(0))
ax[0].set_title('xy')
ax[1].imshow(img.amax(1))
ax[1].set_title('xz')
ax[2].imshow(img.amax(2))
ax[2].set_title('yz')
for x in ax:
    x.set_axis_off()
plt.show()

torch.Size([60, 60, 60, 3])

b. With artifacts

img = neuron_with_artifacts["image"].data.permute(1,2,3,0)

fig, ax = plt.subplots(1,3)
ax[0].imshow(img.amax(0))
ax[0].set_title('xy')
ax[1].imshow(img.amax(1))
ax[1].set_title('xz')
ax[2].imshow(img.amax(2))
ax[2].set_title('yz')
for x in ax:
    x.set_axis_off()
plt.show()

Simulate from existing SWC file

Load real neuron morphology data from an SWC file

labels_dir = os.path.expanduser("~/data/neurotrack_data/simulated_neurons/neuromorpho/swc_trees")
files = [f for x in os.walk(labels_dir) for f in glob(os.path.join(x[0], '*.swc'))]

f_idx = 4
labels_file = files[f_idx]
# load and parse the SWC file data
swc_list = load.swc(labels_file)
sections, sections_graph = load.parse_swc(swc_list)
branches, terminals = load.get_critical_points(swc_list, sections)

loading file: /home/brysongray/data/neurotrack_data/simulated_neurons/neuromorpho/swc_trees/neuromorpho/quinlan/CNG version/KQa4-12-2015-tracing.CNG.swc

Plot SWC raw data

# Create a DataFrame for plotting
data = []
# Iterate through the sections dictionary
for section_id, section_data in sections.items():
    # flatten the section into one list of consecutive points instead of segments (point pairs)
    for i,segment in enumerate(section_data):
        point = segment[0]
        data.append([section_id, point[0].item(), point[1].item(), point[2].item()])
        if i == len(section_data)-1:
            point = segment[1]
            data.append([section_id, point[0].item(), point[1].item(), point[2].item()])
df_sections = pd.DataFrame(data, columns=["section", "x", "y", "z"])

fig = px.line_3d(df_sections, x="x", y="y", z="z", color='section', )
fig.update_layout(scene_aspectmode='data')
fig.show()

Simulate a neuron microscopy image with artifacts from the SWC data

Set random colors for the background and foreground voxels.

random_contrast = True

rng = np.random.default_rng()
neuron_color = np.array([1.0, 1.0, 1.0])
background = np.array([0., 0., 0.])
if random_contrast:
    neuron_color = rng.uniform(size=3)
    neuron_color /= np.linalg.norm(neuron_color)
    background_color = rng.uniform(size=3)
    background_color = background_color / np.linalg.norm(background_color) * 0.01

Draw the neuron image

swc_data = draw.neuron_from_swc(swc_list,
                                width=3,
                                noise=0.00,
                                dropout=False,
                                adjust=True,
                                background_color=background,
                                neuron_color=neuron_color,
                                random_brightness=False,
                                binary=False,
                                rng=rng)

Plot

img = swc_data["image"].data.permute(1,2,3,0)

fig, ax = plt.subplots(1,3, figsize=(18,6))
ax[0].imshow(img.amax(0))
ax[0].set_title('xy')
ax[1].imshow(img.amax(1))
ax[1].set_title('xz')
ax[2].imshow(img.amax(2))
ax[2].set_title('yz')
for x in ax:
    x.set_axis_off()
plt.show()